A method of detecting driver drowsiness state based on multi-features of face

摘要

A real-time method for monitoring driver drowsiness is presented in this paper to prevent potential vehicle accidents. Instead of based on the eye states alone in previous studies, we combined it with the state of driver's mouth to judge whether the driver fatigues, thus solving the traditional challenge of wearing glasses. AdaBoost algorithm is used to detect face region due to its high correct rate. Then the exact positions of driver's eyes and mouth are located according to their geometric features respectively. The method of PATECP (Percentage And Time that Eyelids Cover the Pupils) and PATMIO (Percentage And Time that Mouth Is Open) as well as the new judge rule is used to estimate whether the driver is drowsy. The tests with actual driving video shows that our approach based on eye and mouth features makes the conditions of recognizing the driver's drowsy state wider accurate.